Search NASA⌕ Search

SEARCH · Search NASA

Results for “distributed parallelization”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 469 records · Page 26

A high-performance parallel algorithm for nonnegative matrix factorization

Non-negative matrix factorization (NMF) is the problem of determining two non-negative low rank factors W and H, for the given input matrix A, such that A ≈ WH. NMF is a useful tool for many applications in different domains such as topic modeling in text mining, background separation in video analysis, and community detection in social networks. Despite its popularity in the data mining community, there is a lack of efficient distributed algorithms to solve the problem for big data sets. We propose a high-performance distributed-memory parallel algorithm that computes the factorization by iteratively solving alternating non-negative least squares (NLS) subproblems for W and H. It maintains the data and factor matrices in memory (distributed across processors), uses MPI for interprocessor communication, and, in the dense case, provably minimizes communication costs (under mild assumptions). As opposed to previous implementations, our algorithm is also flexible: (1) it performs well for both dense and sparse matrices, and (2) it allows the user to choose any one of the multiple algorithms for solving the updates to low rank factors W and H within the alternating iterations. We demonstrate the scalability of our algorithm and compare it with baseline implementations, showing significant performance improvements.

Kannan, Ramakrishnan↗

Ion-cyclotron instability in current-carrying Lorentzian (kappa) and Maxwellian plasmas with anisotropic temperatures: A comparative study

Current-driven electrostatic ion-cyclotron instability has so far been studied for Maxwellian plasma with isotropic and anisotropic temperatures. Since satellite-measured particle velocity distributions in space are often better modeled by the generalized Lorentzian (kappa) distributions and since temperature anisotropy is quite common in space plasmas, theoretical analysis of the current-driven, electrostatic ion-cyclotron instability is carried out in this paper for electron-proton plasma with anisotropic temperatures, where the particle parallel velocity distributions are modeled by kappa distributions and the perpendicular velocity distributions are modeled by Maxwellian distributions. Stability properties of the excited ion cyclotron modes and, in particular, their dependence on electron to ion temperature ratio and ion temperature anisotropy are presented in more detail. For comparison, the corresponding results for bi-Maxwellian plasma are also presented. Although the stability properties of the ion cyclotron modes in the two types of plasmas are qualitatively similar, significant quantitative differences can arise depending on the values of {kappa}{sub e} and {kappa}{sub i}. The comparative study is based on the numerical solutions of the respective linear dispersion relations. Quasilinear estimates of the resonant ion heating rates due to ion-cyclotron turbulence in the two types of plasma are also presented for comparison.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Refining HPCToolkit for application performance analysis at exascale

As part of the US Department of Energy’s Exascale Computing Project (ECP), Rice University has been refining its HPCToolkit performance tools to better support measurement and analysis of applications executing on exascale supercomputers. To efficiently collect performance measurements of GPU-accelerated applications, HPCToolkit employs novel non-blocking data structures to communicate performance measurements between tool threads and application threads. To attribute performance information in detail to source lines, loop nests, and inlined call chains, HPCToolkit performs parallel analysis of large CPU and GPU binaries involved in the execution of an exascale application to rapidly recover mappings between machine instructions and source code. To analyze terabytes of performance measurements gathered during executions at exascale, HPCToolkit employs distributed-memory parallelism, multithreading, sparse data structures, and out-of-core streaming analysis algorithms. To support interactive exploration of profiles up to terabytes in size, HPCToolkit’s hpcviewer graphical user interface uses out-of-core methods to visualize performance data. The result of these efforts is that HPCToolkit now supports collection, analysis, and presentation of profiles and traces of GPU-accelerated applications at exascale. These improvements have enabled HPCToolkit to efficiently measure, analyze and explore terabytes of performance data for executions using as many as 64K MPI ranks and 64K GPU tiles on ORNL’s Frontier supercomputer. HPCToolkit’s support for measurement and analysis of GPU-accelerated applications has been employed to study a collection of open-science applications developed as part of ECP. This paper reports on these experiences, which provided insight into opportunities for tuning applications, strengths and weaknesses of HPCToolkit itself, as well as unexpected behaviors in executions at exascale.

Adhianto, Laksono↗

Examining Anomalous Network Performance with Confidence

Variability in network performance is a major obstacle in effectively analyzing the throughput of modern high performance computer systems. High performance interconnection networks offer excellent best-case network latencies; however, highly parallel applications running on parallel machines typically require consistently high levels of performance to adequately leverage the massive amounts of available computing power. Performance analysts have usually quantified network performance using traditional summary statistics that assume the observational data is sampled from a normal distribution. In our examinations of network performance, we have found this method of analysis often provides too little data to understand anomalous network performance. In particular, we examine a multi-modal performance scenario encountered with an Infiniband interconnection network and we explore the performance repeatability on the custom Cray SeaStar2 interconnection network after a set of software and driver updates.

99 GENERAL AND MISCELLANEOUS↗

How fast can one resize a distributed file system?

Efficient resource utilization becomes a major concern as large-scale distributed computing infrastructures keep growing in size. Malleability, the possibility for resource managers to dynamically increase or decrease the amount of resources allocated to a job, is a promising way to save energy and costs. However, state-of-the-art parallel and distributed storage systems have not been designed with malleability in mind. The reason is mainly the supposedly high cost of data transfers required by resizing operations. Nevertheless, as network and storage technologies evolve, old assumptions about potential bottlenecks can be revisited. In this study, we evaluate the viability of malleability as a design principle for a distributed storage system. We specifically model the minimal duration of the commission and decommission operations. To show how our models can be used in practice, we evaluate the performance of these operations in HDFS, a relevant state-of-the-art distributed file system. We show that the existing decommission mechanism of HDFS is good when the network is the bottleneck, but can be accelerated by up to a factor 3 when storage is the limiting factor. We also show that the commission in HDFS can be substantially accelerated. With the highlights provided by our model, we suggest improvements to speed both operations in HDFS. We discuss how the proposed models can be generalized for distributed file systems with different assumptions and what perspectives are open for the design of efficient malleable distributed file systems.

97 MATHEMATICS AND COMPUTING↗

Families of Graph Algorithms: SSSP Case Study

Single-Source Shortest Paths (SSSP) is a well-studied graph problem. Examples of SSSP algorithms include the original Dijkstra’s algorithm and the parallel Δ-stepping and KLA-SSSP algorithms. In this paper, we use a novel Abstract Graph Machine (AGM) model to show that all these algorithms share a common logic and differ from one another by the order in which they perform work. We use the AGM model to thoroughly analyze the family of algorithms that arises from the common logic. We start with the basic algorithm without any ordering (Chaotic), and then we derive the existing and new algorithms by methodically exploring semantic and spatial ordering of work. Our experimental results show that new derived algorithms show better performance than the existing distributed memory parallel algorithms, especially at higher scales.

distributed-memory graph algorithms↗

Berkeley Unified Parallel C (BUPC) Compiler

This program is a portable, open-source, compiler for the UPC language. It is based on the Open64 framework and has extensive support for optimizations. This compiler operates by translating UPC into ANSI/ISO C for compilation by a native compiler and linking with the UPC Runtime Library. This design eases portability to both shared- and distributed-memory parallel architectures. For proper operation it requires the "Berkeley Unified Parallel C (UPC) Runtime Library" and its dependencies.

Iancu, Costin↗

TGE: Machine Learning Based Task Graph Embedding for Large-Scale Topology Mapping

Task mapping is an important problem in parallel and distributed computing. The goal in task mapping is to find an optimal layout of the processes of an application (or a task) onto a given network topology. We target this problem in the context of staging applications. A staging application consists of two or more parallel applications (also referred to as staging tasks) which run concurrently and exchange data over the course of computation. Task mapping becomes a more challenging problem in staging applications, because not only data is exchanged between the staging tasks, but also the processes of a staging task may exchange data with each other. We propose a novel method, called Task Graph Embedding (TGE), that harnesses the observable graph structures of parallel applications and network topologies. TGE employs a machine learning based algorithm to find the best representation of a graph, called an embedding, onto a space in which the task-to-processor mapping problem can be solved. We evaluate and demonstrate the effectiveness of TGE experimentally with the communication patterns extracted from runs of XGC, a large-scale fusion simulation code, on Titan.

Choi, Jong Youl↗

Some Problems and Solutions in Transferring Ecosystem Simulation Codes to Supercomputers

Many computer codes for the simulation of ecological systems have been developed in the last twenty-five years. This development took place initially on main-frame computers, then mini-computers, and more recently, on micro-computers and workstations. Recent recognition of ecosystem science as a High Performance Computing and Communications Program Grand Challenge area emphasizes supercomputers (both parallel and distributed systems) as the next set of tools for ecological simulation. Transferring ecosystem simulation codes to such systems is not a matter of simply compiling and executing existing code on the supercomputer since there are significant differences in the system architectures of sequential, scalar computers and parallel and/or vector supercomputers. To more appropriately match the application to the architecture (necessary to achieve reasonable performance), the parallelism (if it exists) of the original application must be exploited. We discuss our work in transferring a general grassland simulation model (developed on a VAX in the FORTRAN computer programming language) to a Cray Y-MP. We show the Cray shared-memory vector-architecture, and discuss our rationale for selecting the Cray. We describe porting the model to the Cray and executing and verifying a baseline version, and we discuss the changes we made to exploit the parallelism in the application and to improve code execution. As a result, the Cray executed the model 30 times faster than the VAX 11/785 and 10 times faster than a Sun 4 workstation. We achieved an additional speed-up of approximately 30 percent over the original Cray run by using the compiler's vectorizing capabilities and the machine's ability to put subroutines and functions "in-line" in the code. With the modifications, the code still runs at only about 5% of the Cray's peak speed because it makes ineffective use of the vector processing capabilities of the Cray. We conclude with a discussion and future plans.

Skiles, J. W.↗

RADAR: Runtime Asymmetric Data-access Driven Scientific Data Replication

Efficient I/O on large-scale spatiotemporal scientific data requires scrutiny of both the logical layout of the data (e.g., row-major vs. column-major) and the physical layout (e.g., distribution on parallel filesystems). For increasingly complex datasets, hand optimization is a difficult matter prone to error and not scalable to the increasing heterogeneity of analysis workloads. Given these factors, we present a partial data replication system called RADAR. We capture datatype- and collective-aware I/O access patterns (indicating logical access) via MPI-IO tracing and use a combination of coarse-grained and fine-grained performance modeling to evaluate and select optimized physical data distributions for the task at hand. Unlike conventional methods, we store all replica data and metadata, along with the original untouched data, under a single file container using the object abstraction in parallel filesystems. Our system results in manyfold improvements in some commonly used subvolume decomposition access patterns.Moreover, the modeling approach can determine whether such optimizations should be undertaken in the first place.

access pattern↗

TESS

Computing a Voronoi or Delaunay tessellation from a set of points is a core part of the analysis of many simulated and measured datasets. As the scale of simulations and observations surpasses billions of particles, a distributed-memory scalable parallel algorithm is the only feasible approach. The primary contribution of this software is a distributed-memory parallel Delaunay and Voronoi tessellation algorithm based on existing serial computational geometry libraries that automatically determines which neighbor points need to be exchanged among the subdomains of a spatial decomposition. Other contributions include the addition of periodic and wall boundary conditions.

Dmitriy Morozov, TomPeterka↗

Design and implementation of a parallel unstructured Euler solver using software primitives

This paper is concerned with the implementation of a three-dimensional unstructured-grid Euler solver on massively parallel distributed-memory computer architectures. The goal is to minimize solution time by achieving high computational rates with a numerically efficient algorithm. An unstructured multigrid algorithm with an edge-based data structure has been adopted, and a number of optimizations have been devised and implemented to accelerate the parallel computational rates. The implementation is carried out by creating a set of software tools, which provide an interface between the parallelization issues and the sequential code, while providing a basis for future automatic run-time compilation support. Large practical unstructured grid problems are solved on the Intel iPSC/860 hypercube and Intel Touchstone Delta machine. The quantitative effects of the various optimizations are demonstrated, and we show that the combined effect of these optimizations leads to roughly a factor of 3 performance improvement. The overall solution efficiency is compared with that obtained on the Cray Y-MP vector supercomputer.

Das, R.↗

TESS

Computing a Voronoi or Delaunay tessellation from a set of points is a core part of the analysis of many simulated and measured datasets. As the scale of simulations and observations surpasses billions of particles, a distributed-memory scalable parallel algorithm is the only feasible approach. The primary contribution of this software is a distributed-memory parallel Delaunay and Voronoi tessellation algorithm based on existing serial computational geometry libraries that automatically determines which neighbor points need to be exchanged among the subdomains of a spatial decomposition. Other contributions include the addition of periodic and wall boundary conditions.

Dmitriy Morozov, Tom Peterka↗

Gradient-Based Multi-Area Distribution System State Estimation

The increasing distributed and renewable energy resources and controllable devices in distribution systems make fast distribution system state estimation (DSSE) crucial in system monitoring and control. We consider a large multi-phase distribution system and formulate DSSE as a weighted least squares (WLS) problem. We divide the large distribution system into smaller areas of subtree structure, and by jointly exploring the linearized power flow model and the network topology, we propose a gradient-based multi-area algorithm to exactly and efficiently solve the WLS problem. The proposed algorithm enables distributed and parallel computation of the state estimation problem without compromising any performance. Numerical results on a 4,521-node test feeder show that the designed algorithm features fast convergence and accurate estimation results. Comparison with traditional Gauss-Newton method shows that the proposed method has much better performance in distribution systems with a limited amount of reliable measurement. The real-time implementation of the algorithm tracks time-varying system states with high accuracy.

41 EE - Solar Energy Technologies Office (EE-4S)↗

The design and implementation of a parallel unstructured Euler solver using software primitives

This paper is concerned with the implementation of a three-dimensional unstructured grid Euler-solver on massively parallel distributed-memory computer architectures. The goal is to minimize solution time by achieving high computational rates with a numerically efficient algorithm. An unstructured multigrid algorithm with an edge-based data structure has been adopted, and a number of optimizations have been devised and implemented in order to accelerate the parallel communication rates. The implementation is carried out by creating a set of software tools, which provide an interface between the parallelization issues and the sequential code, while providing a basis for future automatic run-time compilation support. Large practical unstructured grid problems are solved on the Intel iPSC/860 hypercube and Intel Touchstone Delta machine. The quantitative effect of the various optimizations are demonstrated, and we show that the combined effect of these optimizations leads to roughly a factor of three performance improvement. The overall solution efficiency is compared with that obtained on the CRAY-YMP vector supercomputer.

Das, R.↗

Planning and Resource Management in an Intelligent Automated Power Management System

Power system management is a process of guiding a power system towards the objective of continuous supply of electrical power to a set of loads. Spacecraft power system management requires planning and scheduling, since electrical power is a scarce resource in space. The automation of power system management for future spacecraft has been recognized as an important R&D goal. Several automation technologies have emerged including the use of expert systems for automating human problem solving capabilities such as rule based expert system for fault diagnosis and load scheduling. It is questionable whether current generation expert system technology is applicable for power system management in space. The objective of the ADEPTS (ADvanced Electrical Power management Techniques for Space systems) is to study new techniques for power management automation. These techniques involve integrating current expert system technology with that of parallel and distributed computing, as well as a distributed, object-oriented approach to software design. The focus of the current study is the integration of new procedures for automatically planning and scheduling loads with procedures for performing fault diagnosis and control. The objective is the concurrent execution of both sets of tasks on separate transputer processors, thus adding parallelism to the overall management process.

Morris, Robert A.↗

The design and implementation of a parallel unstructured Euler solver using software primitives

This paper is concerned with the implementation of a 3D unstructured-grid Euler-solver on massively parallel distributed-memory computer architectures. The goal is to minimize solution time by achieving high computational rates with a numerically efficient algorithm. An unstructured multigrid algorithm with an edge-based data-structure has been adopted, and a number of optimizations have been devised and implemented in order to accelerate the parallel computational rates. The implementation is carried out by creating a set of software tools, which ease the implementation of computational problems on parallel architecture machines by relieving the user of the low-level machine specific issues. The quantitative effect of the various optimizations are demonstrated, and we show that the combined effect of these optimizations leads to roughly a factor of three performance improvement. The overall solution efficiency is compared with that obtained on the CRAY-YMP vector supercomputer.

Das, R.↗

Reducing Communication Costs for Sparse Matrix Multiplication within Algebraic Multigrid

We consider the sequence of sparse matrix-matrix multiplications performed during the setup phase of algebraic multigrid. In particular, we show that the most commonly used parallel algorithm is often not the most communication-efficient one for all of the matrix-matrix multiplications involved. By using an alternative algorithm, we show that the communication costs are reduced (in theory and practice), and we demonstrate the performance benefit for both model (structured) and more realistic unstructured problems on large-scale distributed-memory parallel systems. Our theoretical analysis shows that we can reduce communication by a factor of up to 5.4 for a model problem, and we observe in our empirical evaluation communication reductions of factors up to 4.7 for structured problems and 3.7 for unstructured problems. These reductions in communication translate to run-time speedups of up to factors of 2.3 and 2.5, respectively.

97 MATHEMATICS AND COMPUTING↗